[Ffmpeg-devel-irc] ffmpeg-devel.log.20120219

burek burek021 at gmail.com
Mon Feb 20 02:05:03 CET 2012


[00:00] <Compn> mike at multimedia.cx
[00:00] <Compn> hes not on irc
[02:00] <ubitux> ok i think i automated the coverage stuff
[02:19] <CIA-48> ffmpeg: 03Martin Storsjö 07master * r5302633919 10ffmpeg/libavformat/movenc.c: 
[02:19] <CIA-48> ffmpeg: movenc: Write the unknown duration as 64 bit fields in ismv
[02:19] <CIA-48> ffmpeg: This is required for the files to play back properly in
[02:19] <CIA-48> ffmpeg: windows media player.
[02:19] <CIA-48> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[02:19] <CIA-48> ffmpeg: 03Martin Storsjö 07master * r99a357f4c5 10ffmpeg/libavformat/movenc.c: 
[02:19] <CIA-48> ffmpeg: movenc: Write track durations with all bits set if duration is unknown
[02:19] <CIA-48> ffmpeg: According to 14496-12, the duration should be all 1s if
[02:19] <CIA-48> ffmpeg: the duration is unknown. This is the case if writing a moov
[02:19] <CIA-48> ffmpeg: atom without any samples described in it (e.g. as in ismv files).
[02:19] <CIA-48> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[02:19] <CIA-48> ffmpeg: 03Martin Storsjö 07master * r89baf35996 10ffmpeg/libavcodec/dct-test.c: 
[02:19] <CIA-48> ffmpeg: dct-test: Remove a stray declaration of a nonexistent function
[02:19] <CIA-48> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[02:19] <CIA-48> ffmpeg: 03Martin Storsjö 07master * r0733375059 10ffmpeg/libavcodec/dct-test.c: 
[02:19] <CIA-48> ffmpeg: dct-test: Add the missing ff_ prefix to the altivec functions
[02:19] <CIA-48> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[02:19] <CIA-48> ffmpeg: 03Ronald S. Bultje 07master * rc0994e39d7 10ffmpeg/libavcodec/mpc7.c: mpc7: assign an error level + context to av_log() msg.
[02:19] <CIA-48> ffmpeg: 03Ronald S. Bultje 07master * r9d3050d3e9 10ffmpeg/libavcodec/wmadec.c: 
[02:19] <CIA-48> ffmpeg: wma: don't return 0 on invalid packets.
[02:19] <CIA-48> ffmpeg: Return 0 means "please return the same data again", i.e. it causes an
[02:19] <CIA-48> ffmpeg: infinite loop. Instead, return an error.
[02:19] <CIA-48> ffmpeg: Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
[02:19] <CIA-48> ffmpeg: CC: libav-stable at libav.org
[02:19] <CIA-48> ffmpeg: 03Kostya Shishkov 07master * r6d702dc072 10ffmpeg/libavcodec/proresenc.c: 
[02:19] <CIA-48> ffmpeg: proresenc: force bitrate not to exceed given limit
[02:19] <CIA-48> ffmpeg: Apple ProRes Format Specifications mentions target data size for every frame,
[02:19] <CIA-48> ffmpeg: so make sure frame meets it. This also allows encoder to demand much smaller
[02:19] <CIA-48> ffmpeg: packet sizes for output.
[02:19] <CIA-48> ffmpeg: 03Martin Storsjö 07master * r3eec23f3cd 10ffmpeg/libavformat/ (isom.h mov.c movenc.c): 
[02:19] <CIA-48> ffmpeg: mov: Use defines for trun flags
[02:19] <CIA-48> ffmpeg: Signed-off-by: Martin Storsjö <martin at martin.st>
[02:19] <CIA-48> (64 lines omitted)
[02:21] <CIA-48> ffmpeg: 03Clément BSsch 07master * rf075537561 10ffmpeg/doc/examples/filtering.c: examples/filtering: fix implicit declarations and function mis-usage.
[02:21] <CIA-48> ffmpeg: 03Clément BSsch 07master * rb117c39237 10ffmpeg/doc/examples/filtering.c: examples/filtering: fix deprecated function calls warnings.
[02:21] <CIA-48> ffmpeg: 03Clément BSsch 07master * r2b97be5d57 10ffmpeg/doc/examples/filtering.c: examples/filtering: fix unused variable and return value in open_input_file().
[02:21] <CIA-48> ffmpeg: 03Clément BSsch 07master * r624c413d55 10ffmpeg/doc/examples/Makefile: examples: add -Wall to CFLAGS.
[03:25] <durandal_1707> looks like mono decoding for pam is broken
[03:27] <michaelni> how can i reproduce it ?
[03:27] <durandal_1707> michaelni: easy, pam encoder for mono is not broken
[03:28] <Compn> durandal_1707 : michaelni wants to know command line that fails
[03:28] <Compn> durandal_1707 : so ffmpeg -i file.wav -options...
[03:28] <durandal_1707> create mono file: ffmpeg -i image -pix_fmt monob /tmp/test.pam
[03:28] <Compn> so he can reproduce and fix problem
[03:28] <durandal_1707> play it with ffplay
[03:29] <durandal_1707> file is correctly displayed with display (ImageMagick)
[03:33] <michaelni> is this a regression ?
[03:34] <durandal_1707> dunno
[03:37] <durandal_1707> michaelni: lol, something is weird, monow for xwd is using whole CPU time when using ffplay to display created image, gimp shows it just fine
[03:38] <durandal_1707> michaelni: so i think it is bug in sws, and it is regression
[03:39] <j-b> michaelni: FYI, I have pushed the postproc.git repo on git.v.o
[03:41] <durandal_1707> michaelni: something is really screwed up, ffprobe is using whole CPU too
[03:50] <durandal_1707> michaelni: nah, it is not sws
[04:01] <CIA-48> ffmpeg: 03Michael Niedermayer 07master * re98a76d3c2 10ffmpeg/libavcodec/h264.c: 
[04:01] <CIA-48> ffmpeg: h264: use av_fast_padded_malloc()
[04:01] <CIA-48> ffmpeg: This zeros all the memory once and avoids valgrind warnings.
[04:01] <CIA-48> ffmpeg: alternatively the warnings could be suppressed.
[04:01] <CIA-48> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[04:01] <CIA-48> ffmpeg: 03Paul B Mahol 07master * rb6a08665d6 10ffmpeg/libavformat/img2.c: 
[04:01] <CIA-48> ffmpeg: img2dec: add im32 extension, sunrast with 32bit depth
[04:01] <CIA-48> ffmpeg: Signed-off-by: Paul B Mahol <onemda at gmail.com>
[04:01] <CIA-48> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[04:22] <michaelni> j-b, and others, are there any bugreports about postproc ?
[04:22] <michaelni> i remember some people flaming but i havnt found any bug reports on our tracker
[04:22] <j-b> not that I know of.
[04:22] <j-b> complaints were that it was useless
[04:22] <j-b> not bus
[04:22] <j-b>  bugs
[04:23] <j-b> http://git.videolan.org/?p=libpostproc.git;a=summary . If you do not have write access in it, please tell me.
[04:47] <CIA-48> ffmpeg: 03Michael Niedermayer 07master * r5d468edb2e 10ffmpeg/libavcodec/pamenc.c: 
[04:47] <CIA-48> ffmpeg: pamenc: move allocation down.
[04:47] <CIA-48> ffmpeg: This way the size that is calculated is known and can be used in a following
[04:47] <CIA-48> ffmpeg: commit.
[04:47] <CIA-48> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[04:47] <CIA-48> ffmpeg: 03Michael Niedermayer 07master * rd3f81a6700 10ffmpeg/libavcodec/pamenc.c: 
[04:47] <CIA-48> ffmpeg: pamenc: allocate the correct packetsize for mono.
[04:47] <CIA-48> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[04:54] <durandal_1707> michaelni: that fix is wrong
[04:55] <durandal_1707> encoding is fine, decoding is broken, you just broke encoding
[05:13] <overflow_0f8b> to branch or not to branch? 2 memory writes, or one jump above them ?
[05:16] <durandal_1707> michaelni: actually that 2 last commits could not break pam mono encoding but i dont see what they really fix
[05:32] <michaelni> they allocate a bigger bitstream buffer
[05:33] <durandal_1707> but that should not really be needed
[05:33] <durandal_1707> and that have nothing to do with broken decoding
[05:33] <michaelni> we dont support black&white PAM decoding
[05:34] <durandal_1707> that is funny considering that pbm works
[05:36] <durandal_1707> michaelni: how you got crash?
[05:41] <michaelni> accesses over end of array in valgrind
[05:41] <michaelni> iam not 100% if i tried without valgrind
[05:47] <durandal_1707> that n change for mono does nothing, except increasing packet size
[05:48] <durandal_1707> heh, your change indeed fix bug
[05:48] <durandal_1707> for mono
[05:51] <durandal_1707> but that branch for mono encoding is brain dead
[05:51] <michaelni> i got mono decoding working
[05:52] <durandal_1707> michaelni: make first shure that pam mono encoding is actually correct
[05:53] <michaelni> imgmagick can decode it
[05:53] <michaelni> ive not found anything else yet
[05:53] <durandal_1707> pam = 300k vs  pbm = 37 kb
[05:54] <michaelni> yes iam aware of that
[05:54] <michaelni> when iam happy with my patch ill post it to ffmpeg-dev
[05:54] <durandal_1707> what happened to their brain.....
[05:54] <overflow_0f8b> i need to swap 2 variables if a certain condition is met, but cmov does not have a mem,reg version ;<
[05:58] <durandal_1707> michaelni: there is no nice way to swap bits in byte?
[05:59] <michaelni> if its 1 byte then a LUT can be used as a generic solution
[06:00] <michaelni> if some specific reshuffling is done there might be a nicer way
[06:00] <durandal_1707> lut is what?
[06:00] <michaelni> look up table
[06:00] <durandal_1707> where it is?
[06:01] <michaelni> outputbyte = table[inputbyte]
[06:01] <michaelni> table is 256 entries
[06:01] <durandal_1707> i need to generate one?
[06:02] <michaelni> well if you want to use that suggestion yes, but i dont even know what you are tryig to do exactly so maybe the idea is nit  the best
[06:03] <durandal_1707> i need to swap bits in byte (one byte for now) it is for xbm (X BitMap)
[06:03] <michaelni> overflow_0f8b, one random way is set X to 0 or var0-var1; var0 -= X; var1 += X
[06:04] <durandal_1707> the way how mono plane is stored is incompatible(and brain dead) with sws
[06:04] <michaelni> what does "swap" do exactly?
[06:05] <durandal_1707> 00000101 => 10100000
[06:06] <michaelni> sounds like av_reverse[]
[06:06] <overflow_0f8b> #define __swap__( a1, a2 ) { \
[06:06] <overflow_0f8b>         tmp = (a1) );  \
[06:06] <overflow_0f8b>         (a1) = (a2);    \
[06:06] <overflow_0f8b>         (a2) = tmp;     \
[06:06] <overflow_0f8b> }
[06:06] <overflow_0f8b> swap means exchange
[06:07] <overflow_0f8b> and i can do it, i'm concerned about the jump penalty.
[06:08] <overflow_0f8b> i can do it without jump if i do 2 unnecessary memory writes ;/
[06:08] <overflow_0f8b> memory writes are slow too (compared to nothing)
[06:09] <michaelni> a1 & a2 are in memory ?
[06:09] <overflow_0f8b> yes
[06:09] <overflow_0f8b> it is data
[06:10] <michaelni> cant you just change the algorithm to not need any swap at all ?
[06:11] <michaelni> i mean no conditional swap at least
[06:11] <overflow_0f8b> hmm, i'm open for suggestion on how to swap 2 variables in memory without swapping them
[06:12] <overflow_0f8b> so, if i do swap the 2 variables, it's cool, but if the swap condition is not met, then it does 2 unnecessary memory writes
[06:12] <overflow_0f8b> or one jump above.
[06:13] <michaelni> overflow_0f8b, what is it that you are doing that needs these condiitonal swaps ?
[06:13] <overflow_0f8b> michaelni<< can you suggest another method to let the first operand be the smaller one, and second be the larger ?
[06:14] <overflow_0f8b> #define __minmax0__( min, max )   ( if ( (min) > (max) ) __swap__( (min), (max) ) )
[06:14] <durandal_1707> michaelni: thanks
[06:14] <michaelni> durandal_1707, np
[06:15] <michaelni> overflow_0f8b, what is that minmax0 used in ?
[06:15] <overflow_0f8b> a sorting algorithm?
[06:16] <overflow_0f8b> michaelni<< btw i believe a table is not needed for the bitswapping.
[06:17] <michaelni> use radix sort
[06:19] <michaelni> that is unless the array is very small or nearly sorted already
[06:34] <overflow_0f8b> michaelni<< will you recommend a table for swapping all 64 bits in a 64 bit variable?
[06:38] <michaelni> when you want to bit reverse a 64bit variable, its probably faster without a table
[06:39] <overflow_0f8b> and you say it will be faster to read a table for 8 bit?
[06:39] <overflow_0f8b> then using bit magic?
[06:40] <michaelni> i dont really know where what is faster, best you try both
[14:50] Action: Daemon404 swears michaelni may as well be on north american time, given the hours he is awake...
[14:52] <durandal_1707> ffprobe/ffmpeg/ffplay still hangs for me on xwd
[14:52] <Daemon404> infinite loop?
[14:53] <durandal_1707> it takes to much time so yes it is probably (nearly) infinite
[14:54] <durandal_1707> after say few seconds I just use ctrl+c
[14:54] <Daemon404> it could just be incredibly inefficient
[14:54] <durandal_1707> it hangs before decoding
[15:03] <durandal_1707> lol, removed my experimental encoder/decoder and problem is gone....
[15:04] <durandal_1707> heh, pebkac
[15:29] <CIA-48> ffmpeg: 03Reimar Döffinger 07master * r3fa6d20533 10ffmpeg/libavcodec/h264_parser.c: 
[15:29] <CIA-48> ffmpeg: H264 parser: enable header parsing/frame type detection for is_avc.
[15:29] <CIA-48> ffmpeg: Signed-off-by: Reimar Döffinger <Reimar.Doeffinger at gmx.de>
[16:54] <ubitux> oh nice, the two valgrind boxes in the green
[17:07] <CIA-48> ffmpeg: 03Michael Niedermayer 07master * rd9b3097ba2 10ffmpeg/libavcodec/ (pnm.c pnmdec.c): 
[17:07] <CIA-48> ffmpeg: pamdec: fix mono support.
[17:07] <CIA-48> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[19:41] <CIA-48> ffmpeg: 03Benjamin Larsson 07master * r7f215e3e72 10ffmpeg/libavcodec/wmalosslessdec.c: Use fixed-length array in revert_mclms()
[19:41] <CIA-48> ffmpeg: 03Mashiat Sarker Shakkhar 07master * ra520a9946f 10ffmpeg/libavcodec/wmalosslessdec.c: Perform inter-channel decorr. only if both channels are coded
[19:41] <CIA-48> ffmpeg: 03Michael Niedermayer 07master * r04f7eee709 10ffmpeg/: 
[19:41] <CIA-48> ffmpeg: Merge remote-tracking branch 'shariman/wmall'
[19:41] <CIA-48> ffmpeg: * shariman/wmall:
[19:41] <CIA-48> ffmpeg:  Perform inter-channel decorr. only if both channels are coded
[19:41] <CIA-48> ffmpeg:  Use fixed-length array in revert_mclms()
[19:41] <CIA-48> ffmpeg: Merged-by: Michael Niedermayer <michaelni at gmx.at>
[22:12] <CIA-48> ffmpeg: 03Michael Niedermayer 07master * r1060e9ce54 10ffmpeg/ (cmdutils.c cmdutils.h cmdutils_common_opts.h): 
[22:12] <CIA-48> ffmpeg: cmdutils: add -cpuflags.
[22:12] <CIA-48> ffmpeg: This allows globally forcing specific cpuflags (or lack thereof)
[22:12] <CIA-48> ffmpeg: Useful for debugging and benchmarking
[22:12] <CIA-48> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[22:12] <CIA-48> ffmpeg: 03Michael Niedermayer 07master * rb240866c5b 10ffmpeg/libavcodec/wmalosslessdec.c: 
[22:12] <CIA-48> ffmpeg: wmalossless: remove tab
[22:12] <CIA-48> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[22:12] <CIA-48> ffmpeg: 03Michael Niedermayer 07master * r886b40b94a 10ffmpeg/libavcodec/wmalosslessdec.c: 
[22:12] <CIA-48> ffmpeg: wmalossless: Switch to new audio API
[22:12] <CIA-48> ffmpeg: Partially fixes Ticket1000
[22:12] <CIA-48> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[22:12] <CIA-48> ffmpeg: 03Michael Niedermayer 07master * rf4f386dd00 10ffmpeg/libavcodec/wmalosslessdec.c: 
[22:12] <CIA-48> ffmpeg: wmalossless: Fix infinite loop.
[22:12] <CIA-48> ffmpeg: Fixes more of Ticket1000
[22:12] <CIA-48> ffmpeg: Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
[00:00] --- Mon Feb 20 2012


More information about the Ffmpeg-devel-irc mailing list